Re: [PATCH] cpufreq: intel_pstate: Override parameters if HWP forced by BIOS

[Rafael J. Wysocki]

On Thu, Sep 9, 2021 at 6:12 PM Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote:
>
> On Thu, Sep 9, 2021 at 3:20 PM Doug Smythies <dsmythies@xxxxxxxxx> wrote:
> >
> > On Thu, Sep 9, 2021 at 4:18 AM Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote:
> > >
> > > On Thu, Sep 9, 2021 at 8:52 AM Srinivas Pandruvada
> > > <srinivas.pandruvada@xxxxxxxxxxxxxxx> wrote:
> > > >
> > > > On Wed, 2021-09-08 at 20:48 -0700, Doug Smythies wrote:
> > > > > If HWP has been already been enabled by BIOS, it may be
> > > > > necessary to override some kernel command line parameters.
> > > > > Once it has been enabled it requires a reset to be disabled.
> > > > >
> > > > > Signed-off-by: Doug Smythies <dsmythies@xxxxxxxxx>
> > > > > ---
> > > > >  drivers/cpufreq/intel_pstate.c | 22 ++++++++++++++++------
> > > > >  1 file changed, 16 insertions(+), 6 deletions(-)
> > > > >
> > > > > diff --git a/drivers/cpufreq/intel_pstate.c
> > > > > b/drivers/cpufreq/intel_pstate.c
> > > > > index bb4549959b11..073bae5d4498 100644
> > > > > --- a/drivers/cpufreq/intel_pstate.c
> > > > > +++ b/drivers/cpufreq/intel_pstate.c
> > > > > @@ -3267,7 +3267,7 @@ static int __init intel_pstate_init(void)
> > > > >                  */
> > > > >                 if ((!no_hwp && boot_cpu_has(X86_FEATURE_HWP_EPP)) ||
> > > > >                     intel_pstate_hwp_is_enabled()) {
> > > > > -                       hwp_active++;
> > > > > +                       hwp_active = 1;
> > > > Why this change?
> > >
> > > I think hwp_active can be changed to bool and then it would make sense
> > > to update this line.
> > >
> > > > >                         hwp_mode_bdw = id->driver_data;
> > > > >                         intel_pstate.attr = hwp_cpufreq_attrs;
> > > > >                         intel_cpufreq.attr = hwp_cpufreq_attrs;
> > > > > @@ -3347,17 +3347,27 @@ device_initcall(intel_pstate_init);
> > > > >
> > > > >  static int __init intel_pstate_setup(char *str)
> > > > >  {
> > > > > +       /*
> > > > > +        * If BIOS is forcing HWP, then parameter
> > > > > +        * overrides might be needed. Only print
> > > > > +        * the message once, and regardless of
> > > > > +        * any overrides.
> > > > > +        */
> > > > > +       if(!hwp_active
> > > > This part of code is from early_param, Is it possible that
> > > > hwp_active is not 0?
> > >
> > > Well, it wouldn't matter even if it were nonzero.  This check is just
> > > pointless anyway.
> > >
> > > > > && boot_cpu_has(X86_FEATURE_HWP))
> > > > > +               if(intel_pstate_hwp_is_enabled()){
> > >
> > > This should be
> > >
> > > if (boot_cpu_has(X86_FEATURE_HWP) && intel_pstate_hwp_is_enabled()) {
> >
> > Disagree.
> > This routine gets executed once per intel_pstate related grub command
> > line entry. The purpose of the "if(!hwp_active" part is to prevent the
> > printing of the message to the logs multiple times.
>
> Ah OK.  Fair enough.
>
> You can do all of the checks in one conditional, though.  They will be
> processed left-to-right anyway.
>
> But then it would be good to avoid calling
> intel_pstate_hwp_is_enabled() multiple times if it returns false.
>
> And having said all that I'm not sure why you are trying to make
> no_hwp depend on !hwp_active?  I will not be taken into account anyway
> if intel_pstate_hwp_is_enabled() returns 'true'?
>
> So if no_hwp is covered regardless, you may move the
> intel_pstate_hwp_is_enabled() inside the no_load conditional.
>
> Alternatively, and I would do that, intel_pstate_hwp_is_enabled()
> could be evaluated earlier in intel_pstate_init() and if it returned
> 'true', both no_load and no_hwp would be disregarded.

Something like the attached, for the record.
---
 drivers/cpufreq/intel_pstate.c |   16 +++++++++++-----
 1 file changed, 11 insertions(+), 5 deletions(-)

Index: linux-pm/drivers/cpufreq/intel_pstate.c
===================================================================
--- linux-pm.orig/drivers/cpufreq/intel_pstate.c
+++ linux-pm/drivers/cpufreq/intel_pstate.c
@@ -3205,11 +3205,15 @@ static int __init intel_pstate_init(void
 	if (boot_cpu_data.x86_vendor != X86_VENDOR_INTEL)
 		return -ENODEV;
 
-	if (no_load)
-		return -ENODEV;
-
 	id = x86_match_cpu(hwp_support_ids);
 	if (id) {
+		bool hwp_forced = intel_pstate_hwp_is_enabled();
+
+		if (hwp_forced)
+			pr_info("HWP enabled by BIOS\n");
+		else if (no_load)
+			return -ENODEV;
+
 		copy_cpu_funcs(&core_funcs);
 		/*
 		 * Avoid enabling HWP for processors without EPP support,
@@ -3219,8 +3223,7 @@ static int __init intel_pstate_init(void
 		 * If HWP is enabled already, though, there is no choice but to
 		 * deal with it.
 		 */
-		if ((!no_hwp && boot_cpu_has(X86_FEATURE_HWP_EPP)) ||
-		    intel_pstate_hwp_is_enabled()) {
+		if ((!no_hwp && boot_cpu_has(X86_FEATURE_HWP_EPP)) || hwp_forced) {
 			hwp_active++;
 			hwp_mode_bdw = id->driver_data;
 			intel_pstate.attr = hwp_cpufreq_attrs;
@@ -3236,6 +3239,9 @@ static int __init intel_pstate_init(void
 			goto hwp_cpu_matched;
 		}
 	} else {
+		if (no_load)
+			return -ENODEV;
+
 		id = x86_match_cpu(intel_pstate_cpu_ids);
 		if (!id) {
 			pr_info("CPU model not supported\n");